Path: news.mitre.org!blanket.mitre.org!agate!hammer.uoregon.edu!netnews.nwnet.net!news.microsoft.com!news From: "kyleb" Newsgroups: comp.os.ms-windows.programmer.nt.kernel-mode Subject: Re: NDISwith layered drivers - q156655 and problems Date: Thu, 6 Nov 1997 09:52:57 -0800 Organization: microsoft corp. Lines: 60 Message-ID: <63t05t$5g3@news.microsoft.com> References: <01bceaa0$00a03ea0$6000a8c0@socrates.acu.rl.ac.uk> NNTP-Posting-Host: 172.31.203.213 X-Newsreader: Microsoft Outlook Express 4.71.1008.4 X-MimeOle: Produced By Microsoft MimeOLE Engine V4.71.1008.4 are you creating a device object in your miniport and sending irps to that? where do you fix up the dispatch handlers? this isn't a supported method but if i were going to do something like this i would make sure that i swaped the dispatch handlers in the DriverEntry routine of my miniport AFTER NdisMRegisterMiniport returned. i would also have a reliable way to identify a device object that i allocated. something that couldn't be confused with a device object that ndis allocated. then when a dispatch handler was called i would verify the devobj and if it wasn't mine i would pass it on to the ndis dispatch handlers ( i would have saved those by the way). at least thats where i would start.. i would also never make the assumption that any of this would work on NT 5.0.... ;) - The opinions expressed in this message are my own personal views and do not reflect the official views of Microsoft Corporation please respond to newsgroup. Do not send mail. christopher john adams wrote in article <01bceaa0$00a03ea0$6000a8c0@socrates.acu.rl.ac.uk>... >I have written a layered miniport driver which runs over an ATM driver. The >ATM driver is not an ndis driver but communicates by IRPs with the miniport >driver and directly with the IO manager for applications. The ATM low level >driver works fine and is used for multi-client video and tv traffic with >quite heavy work loads. > >The ndis driver also works but NT has a tendency to crash after 1-2 hrs. >The crash does not happen in my code (or even related thread) but in NDIS >which when traced back is due to a pageable ndis routine being called when >the code is not paged in. (we used softice to trace this back - although we >did not have the kernel symbol table). > > > >After a couple of weeks of trying to find what I was doing to upset ndis I >came across q156655 which describe problems with memory leaks and >IRQ_NOT_LESS_OR_EQUALs in ndis. These seem to describe my symptoms pretty >well so I tried the hot fix. Unfortunately this made matters much worse as >on the second IRP Iocompletion, NT falls over in the kernel with an INLOE >at 80115ccc. > >(I also locked ndis pages. This prevented the pageing problem but caused >problems with the filing system (FAT) later on) > >My suspition is that I have a related problem to the one described in >q156655 but it is slightly different as I am using a ordinary low level >driver rather than an NDIS layered driver. > > >Can anyone help or comment on my difficulties. Usually I love this sort of >thing but time is pressing on! > >Cheers Chris > >chris@buck.ac.uk